Research On Reconstruct Of Full Focus Image

With the rapid development of multimedia and computer technology, image processing becomes more and more widely used. Blind image restoration is a great part of image processing, which refers to the technology field that study how to restore the original image from the observed image without a sufficient prior knowledge about the degradation process. The general sequence of blind image restoration is: firstly, model the degradation processing in mathematical form, and secondly, solve this inverse problem to get an estimation of original image.There are many methods of image restoration, but most of which are just discussed in theory, fewer in practice. As the paper adopted the methods of rapid Fourier arithmetic and transform from enormous iterative matrix in dimensional domain to iterative matrix can be accepted in personal computer in frequency domain, and in the iterative processing of dimensional domain, product of matrix can be turn into product of corresponding element in frequency domain, then, EMS memory occupied in processing and run time has been reduced out and away, constringency of iterative course is accelerated with the way of regularized iterative image restoration approach in the frequency.In this thesis, we focus on the research of blind image restoration, which consists of the following parts:(1) The image models and methods of blind image restoration are presented. Besides, the measured standards have been introduced.(2) The wiener filter and Constrained least squares algorithm are analyzed. A new and faster image blind restoration methods has been put forward.(3) According to the law of the energy conservation. The algorithm of full focus image reconstruct has been proposed. When we can not know blurred radii of defocus blurred image, in the processing of searching optimal iterative image, we may think defocus radii is a variable, then calculational capacity is increased greatly and time of constringency is prolonged, but time of once iterative account isn't long, so the arithmetic is possible that the best value of defocus blurred radii has been searched as a variable.(4) NAS-RIF (Nonnegativity and Support Constrains Recursive Inverse Filtering). The basic idea of NAS-RIF is given and by integrating the idea of an improved algorithm is given. Then effectiveness of this algorithm is proved by many simulations.